accelerate
Model trainer
A tool to simplify training and deployment of PyTorch models on various devices and configurations
🚀 A simple way to launch, train, and use PyTorch models on almost any device and distributed configuration, automatic mixed precision (including fp8), and easy-to-configure FSDP and DeepSpeed support
8k stars
97 watching
968 forks
Language: Python
last commit: 6 days ago
Linked from 3 awesome lists
ethicalml/awesome-production-machine-learning
bharathgs/awesome-pytorch-list
kelvins/awesome-mlopsRelated projects:
| Repository | Description | Stars |
|---|---|---|
 nvidia/apex | Tools for streamlined mixed precision and distributed training in PyTorch | 8,407 |
 pytorch/pytorch | A Python library providing tensors and dynamic neural networks with strong GPU acceleration | 83,959 |
| pytorch/torchtitan | A native PyTorch library for large-scale language model training with distributed training capabilities | 2,615 |
| huggingface/optimum | A toolkit providing optimization tools and hardware acceleration for training and inference of machine learning models | 2,572 |
| pytorch/serve | A tool for serving and scaling PyTorch models in production environments | 4,217 |
 facebookresearch/fairscale | A PyTorch extension library that provides high-performance and large-scale training techniques. | 3,193 |
 rtqichen/torchdiffeq | A PyTorch implementation of differentiable ODE solvers with GPU support and efficient backpropagation | 5,588 |
| pytorch/ignite | A high-level library to help with training and evaluating neural networks in PyTorch | 4,526 |
 zllrunning/face-parsing.pytorch | Implementation of a face parsing model using PyTorch and a modified BiSeNet architecture. | 2,317 |
 amdegroot/ssd.pytorch | An implementation of a deep learning-based object detection system in PyTorch. | 5,146 |
 catalyst-team/catalyst | A PyTorch framework for accelerating deep learning research and development by focusing on reproducibility, rapid experimentation, and codebase reuse. | 3,295 |
 moskomule/senet.pytorch | An implementation of Squeeze-and-Excitation Networks (SE-Nets) for deep learning image classification tasks | 2,285 |
| huggingface/peft | An efficient method for fine-tuning large pre-trained models by adapting only a small fraction of their parameters | 16,437 |
 jwyang/faster-rcnn.pytorch | An object detection implementation built on top of PyTorch, supporting multi-image batch training and multiple GPUs. | 7,693 |
| pytorch/glow | A compiler and execution engine for neural networks that generates optimized code for hardware accelerators | 3,235 |